Compile-time reflection will remove countless of lines of bloat from real C++ code bases due to eliminating the need to manually write formatting, hashing and serialisation implementations for classes.

…10 years from now

If not more; the words "modern" and "C++11" are still used in conjunction despite the fact that 2014 was a long time ago.

All programming languages either die, or become bloated, as any software product.

I bet a Fortran 77 developer will think the same of Fortran 2023, a COBOL 60 developer of COBOL 2023, a K&R C developer from C23, a 1975 Scheme developer from R7RS, a Python 1.0 developer from Python 3.13,... even Go 1.0 developers from 1.24 with generics, generators,...

One trick that can help you here is Rust's Editions (and the proposed but never implemented C++ Epochs).

This lets the language throw away bad ideas, without throwing away the code people wrote in the era when we didn't realise that's a bad idea.

Kind of, because epochs only cover grammar changes and some kinds of semantic changes.

It doesn't cover standard library, scenarios with binary libraries doing cross calls across epochs and several others.

I still don't see it any different from a language version switch like in many other languages, specially in JVM and .NET ecosystem.

I am not sure I necessarily agree with that. It matters what means of extending the language itself the language provides and thereby enables the users of the language to add to it, in form of libraries, that don't have to be part of the standard distribution of the language, but can still be reached for by anyone who wants to select them.

Languages with good macro systems have the upper hand in that regard.